What is Prompt Tuning and Prompt Engineering Explained #promptengineering #prompttuning

This are the Techniques which allows us to improve the performance of Pretrained LLMs -
1. Fine Tuning
2.Prompt Tuning
3.Prompt Engineering

Until recently the best way was Fine Tuning but there's a simpler , far more energy efficient technique Known as Prompt Tuning .

In Prompt tuning the best clues are frontend Prompts that are fed to your AI model to give it task - specific context , the prompts can be extra words introduced by a human or more commonly an AI generated number introduced into the model's embedding layer to guide the model towards a desired prediction

Soft Prompts outperforms human engineered prompts and they are unrecognizable to the human eye .
Each prompt consist of embedding or string of numbers that distill knowledge from large model .

Drawback of Prompt Tuning
- Lack of interpretability that means the AI discovers Prompts and optimize it for a given task but often can't explain why it chose those embeddings .

PT is a game changer in Variety of areas like multitask and continual learning .

We use Prompt Engineering to train a model to perform a specialized task with just a single prompt introduced to inference time without needing the model to be retrained .

#ai #prompt #promptengineering #prompttuning #ailearning #technology #llm #pretrainedmodels